When should you operate the extractor yourself?
The Python library can run in your environment. Use it when you have a maintainer for the browser and extraction process. The hosted service is a separate option with its own usage and terms. Check those separately before choosing.
What should you keep beside the Markdown?
Keep the original location, retrieval date and the extraction settings you used. Compare headings and tables with the original before treating the text as a complete representation. Use sources you own or have permission to process.
What costs remain for the open-source option?
The repository identifies the crawler as Apache 2.0. Budget for the machine, browser runtime and maintenance, plus any extra service you choose to connect. The hosted product uses a separate usage allowance.
Which lines belong in an extracted help article?
Draw your own fictional help page with a breadcrumb, two instructions and a footer. Highlight the two instructions the text copy needs.
- Keep heading contextPair each instruction with its heading and original address.
- Mark missing contentImagine a collapsed step is absent from the copy. Require a human check before using that copy to answer a question.